论文部分内容阅读
本文用修正后的高斯－牛顿方法评价非恒定地下水流几种参数估算法的精度。计算敏感系数矩阵的方法有三种：影响系数法、敏感方程法和变分法。后两种方法在计算中受初始参数和附加样本误差的影响最小。当时间序列数据中存在随机或系统误差时，参数预测结果益发受初始导水系数矢量形式的影响。修正后的高斯－牛顿方法因其计算简单、收敛速度快而具有明显的优越性。特别是当参数数目不太多时，这种优越性更为明显。
In this paper, the modified Gauss-Newton method is used to evaluate the precision of several parameters estimation methods for unsteady groundwater flow. There are three ways to calculate the matrix of sensitive coefficients: the influence coefficient method, the sensitive equation method and the variational method. The latter two methods are least affected by initial parameters and additional sample errors in the calculation. When random or systematic errors exist in the time-series data, the prediction of the parameters is affected by the vector form of the initial hydraulic conductivity. The modified Gauss-Newton method has obvious advantages because of its simple calculation and fast convergence speed. Especially when the number of parameters is not too much, this advantage is more obvious.